Re: [O] [RFC] Org linting library

[Andreas Leha]

Hi Rainer,

Rainer M Krug <address@hidden> writes:
> Rainer M Krug <address@hidden> writes:
>
>> Rainer M Krug <address@hidden> writes:
>>
>>> Andreas Leha <address@hidden> writes:
>>>
>>>> Hi Rainer,
>>>>
>>>> Rainer M Krug <address@hidden> writes:
>>>>> Nicolas Goaziou <address@hidden> writes:
>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> The following library implements linting for Org syntax. The sole public
>>>>>> function is `org-lint', which see.
>>>>>>
>>>>>> Internally, the library defines a new structure: `org-lint-checker',
>>>>>> with the following slots:
>>>>>>
>>>>>>   - NAME: Unique check identifier, as a symbol. The check is done
>>>>>>     calling the function `org-lint-NAME' with one mandatory argument,
>>>>>>     the parse tree describing the current Org buffer. Such function
>>>>>>     calls are wrapped within a `save-excursion' and point is always at
>>>>>>     `point-min'. Its return value has to be an alist (POSITION MESSAGE)
>>>>>>     when POSITION refer to the buffer position of the error, as an
>>>>>>     integer, and MESSAGE is a strings describing the error.
>>>>>>
>>>>>>   - DESCRIPTION: Summary about the check, as a string.
>>>>>>
>>>>>>   - CATEGORIES: Categories relative to the check, as a list of symbol.
>>>>>>     They are used for filtering when calling `org-lint'. Checkers not
>>>>>>     explicitly associated to a category are collected in the `default'
>>>>>>     one.
>>>>>>
>>>>>>   - TRUST: The trust level one can have in the check. It is either `low'
>>>>>>     or `high', depending on the heuristics implemented and the nature of
>>>>>>     the check. This has an indicative value only and is displayed along
>>>>>>     reports.
>>>>>>
>>>>>> All checks have to be listed in `org-lint--checkers'.
>>>>>>
>>>>>> Results are displayed in a special "*Org Lint*" buffer with a dedicated
>>>>>> major mode, derived from `tabulated-list-mode'. In addition to the usual
>>>>>> key-bindings inherited from it, "C-j" displays problematic line reported
>>>>>> under point and "RET" jumps to it.
>>>>>>
>>>>>> Checks currently implemented are:
>>>>>>
>>>>>>   - duplicates CUSTOM_ID properties
>>>>>>   - duplicate NAME values
>>>>>>   - duplicate targets
>>>>>>   - duplicate footnote definitions
>>>>>>   - orphaned affiliated keywords
>>>>>>   - obsolete affiliated keywords
>>>>>>   - missing language in src blocks
>>>>>>   - NAME values with a colon
>>>>>>   - wrong header arguments in src blocks
>>>>>>   - misuse of CATEGORY keyword
>>>>>>   - "coderef" links with unknown destination
>>>>>>   - "custom-id" links with unknown destination
>>>>>>   - "fuzzy" links with unknown destination
>>>>>>   - "id" links with unknown destination
>>>>>>   - links to non-existent local files
>>>>>>   - special properties in properties drawer
>>>>>>   - obsolete syntax for PROPERTIES drawers
>>>>>>   - missing definition for footnote references
>>>>>>   - missing reference for footnote definitions
>>>>>>   - non-footnote definitions in footnote section
>>>>>>   - probable invalid keywords
>>>>>>   - invalid blocks
>>>>>>   - probable incomplete drawers
>>>>>>   - obsolete QUOTE section
>>>>>>
>>>>>> Since it relies on lexical binding, `pcase' and `string-prefix-p', it
>>>>>> cannot be added to Org 8.3, but can make it into Org 8.4, if deemed
>>>>>> useful enough.
>>>>>>
>>>>>
>>>>> This sounds very interesting and I would like to try it out. I
>>>>> understand that it can't be put into master, but could it be put into a
>>>>> branch?
>>>>>
>>>>> This would make testing a bit easier.
>>>>>
>>>>
>>>> It is.  The branch is called `wip-lint'.
>>>
>>> Thanks (also to Nicolas) - I found it. Just expected the branch to be
>>> tracked automatically.
>>>
>>> This is really brilliant!
>>>
>>> But I now get a message in one .org file:
>>>
>>> ,----
>>> | Org linting process starting...
>>> | Search failed: "^[        ]*#\\+NAME: +tab:sensVar"
>>> `----
>>>
>>> and no results.
>>>
>>> Works in other .org files.
>>>
>>> This one is rather long (11570 lines) and many code blocks.
>>>
>>> Just let me know how I can trace down where this is coming from and what
>>> the message tells me.
>>
>> It seems that the error comes from the fact that ~#+LABEL: sensVar~ was
>> defined twice.
>>
>> Renaming these results in working linting.
>
> OK - please ignore this last comment.
>
> There is an example where I get the error:
>
> * Fitting the kernel to the data
> The parameter which will be fitted can be found in Table [[tab:fitInitial]]
>
> #+CAPTION: Variables used for the initial fit of the wind profile using the 
> function and the initial values.
> #+LABEL: tab:fitInitial
> | variable           | initial value | remark                                 
>           |
> |--------------------+---------------+--------------------------------------------------|
>
> The cause is the link [[tab:initial]] If I remove everything below and
> including the line #+CAPTION the linting works.
>
> ,----
> |      2 high  Unknown fuzzy location "tab:fitInitial"
> `----
>

Untested - but should that not be
#+NAME: tab:fitInitial
rather than #+LABEL?

Cheers,
Andreas